THE PENNSYLVANIA STATE UNIVERSITY Office of Telecommunications GUIDE TO THE PENN STATE DATA BACKBONE Version 1 August 10, 1989 TABLE OF CONTENTS 1. INTRODUCTION 1 1.1 Purpose of this Document 1 1.2 Scope of this Document 1 1.3 Structure of this Document 1 1.4 Comments and Corrections 2 2. NETWORK DESCRIPTION 2 2.1 Environment 3 2.2 General Description 3 2.3 Hardware 4 2.4 Protocol 4 3. CONNECTION 5 3.1 Physical Connection 5 3.1.1 Fiber Links 5 3.1.2 Gateway Access Points 6 3.1.3 Local Network Attachment 6 3.2 Protocol Implementation 6 3.2.1 Features of TCP/IP 7 3.2.2 PCs 8 3.2.3 Asynchronous Terminals 9 4. ACCESS PROCEDURES 9 4.1 On Campus 10 4.2 Off Campus 12 5. SUPPORT 13 5.1 Ordering a Data Backbone Connection 13 5.2 New User Requirements 14 5.3 Reporting Data Backbone Trouble 15 6. APPENDICES 16 Host Table Format 16 PSU Networking Glossary 19 General Networking Glossary 20 Relevant OTC Policy Statements 34 1. INTRODUCTION 1.1 Purpose of this Document This document is written to describe the PSU Data Backbone in terms of its implementation and use. It outlines the network architecture, the network software protocols, the conditions for attachment to the network and considerations for use of the network. It is intended to be a guide for Penn State faculty and/or staff who are planning to initiate use of the Data Backbone or who wish to extend or enhance their current use. 1.2 Scope of this Document This document is meant to be specific to the PSU Data Backbone networking environment. It is not meant to be a general tutorial on data networks. It does not describe other data transmission networks operated by OTC (see Section 2.1) except to explain the Data Backbone interface to such networks. It also does not attempt to describe the use of particular LANs which may be attached to the Data Backbone except to specify the LAN interface that is required for such an attachment. Nor does this document describe private networks which may be installed, operated or maintained by a particular organization (such as a college, department or administrative office). 1.3 Structure of this Document Section 2, NETWORK DESCRIPTION, defines the Data Backbone in the overall context of telecommunications at PSU. It describes the telecommunications environment and gives a general description of the Data Backbone. It outlines briefly the hardware and software elements of the network and shows a simple graphic outline of the network structure. Section 3, CONNECTION, defines the Data Backbone in terms of means of connection to the network. Physical connection is outlined in terms of the fiber-optic cable link required, the hardware router/gateway to which the LAN must attach and the types of LANs which can be so connected. Software connection is described in terms of TCP/IP protocol implementation for PCs and data terminals and some examples of existing LAN implementation strategies in use at PSU are given. Section 4, ACCESS PROCEDURES, outlines procedures that give user access to computing resources via the Data Backbone. Internet address and domain-name structure are discussed. An outline is given of TCP/IP procedures for interactive and file transfer use of the network with examples specific to the PSU environment. Use of the cisco terminal server is discussed, both for access to non-TCP/IP host services and for RS-232 data switch connection to the Data Backbone. This section concludes with a description of Data Backbone connection to off-campus (NSFnet, PREPnet etc.) Internet services. Section 5, SUPPORT, Outlines the procedure for ordering a Data Backbone connection from OTC. It outlines Data Backbone cabling policy and defines the demarcation point for LAN attachment to the network. It describes some considerations for users in terms of coordinating LAN management with OTC and gives the OTC Trouble Line No. Section 6, APPENDICES. Appendix 1 outlines Internet name server host table format and gives a source for further reference information on this subject. Appendices 2 and 3 are glossaries of networking terminology for use in clarifying discussions of network architecture. Appendix 2 is for PSU-specific networking. Appendix 3 is for general networking. Appendix 4 contains OTC policy statements on Data Backbone Attachments, on Routing DECnet Packets on the Data Backbone, on Intra-building Cabling and on Inter-building Cabling. 1.4 Comments and Corrections Any comments on or corrections to this document or requests for supplementary information can be directed to Neal K. Todd at Office of Telecommunications Data Services in Pine Cottage. The telephone no. is 865-2404. The electronic mail address is N6T at PSUVM. 2. NETWORK DESCRIPTION PSU maintains a large set of data, video and voice telecommunications networking resources in order to assist the separate campuses and departments of the University in functioning as an integrated whole. The subset of these resources which concerns itself with data transmission forms an interconnected system of networks of which the Data Backbone is a major component. The Data Backbone is implemented with modern high-speed technology using Proteon fiber-optic hardware and TCP/IP data communications protocols. 2.1 Environment OTC maintains and operates an interconnected set of data communications networks in order to provide access to data resources to the separate campuses and departments which constitute the University. The State-Wide Network (as defined in Appendix 2) is operated to interconnect the regional campuses and University Park. The Data-Switch Network (see Appendix 2 -- Switched Network) connects RS-232 asynchronous devices to certain host data resources. The Data Backbone is a high-speed fiber-optic network which operates to interconnect departmental networks at University Park. 2.2 General Description A campus-wide high-speed data communications network has been developed at University Park. The network supports interconnectivity of LANs, hosts and workstations and has the high-speed volume appropriate for growth of information processing. This network functions as a data backbone in that it interconnects numerous departmental networks. It connects a number of diverse systems comprised of many types of computer hardware and software. These systems contain a multiplicity of data resources for academic and research use. Connection at high-speed is important to many users because it allows them to move large files rapidly throughout the network to better support research activities. The Data Backbone includes all devices and cabling necessary to support data transmission between the user's local network and other local networks (including host centers). It includes any inter- and intra-building cabling (and the associated conduit space). It includes the software needed to enable communications between the Data Backbone routers. The Data Backbone technology is implemented in a pair of high-speed fiber-optic token ring communications networks consisting of collections of hardware elements such as fiber transmission links, fiber-driver units, intelligent router-gateway nodes with interfaces for LAN attachment and various network file servers and monitoring stations. The network elements operate under the control of the TCP/IP suite of software telecommunications protocols which coordinate the flow of user traffic and allow network management. The Data Backbone is centrally managed and operated by the Office of Telecommunications (OTC). 2.3 Hardware The hardware technology of the Data Backbone establishes the network configuration as a double ring. The inner ring is a Proteon ProNET-80. ProNET-80 is an 80 megabit/sec star-shaped fiber-optic token ring network. The inner ring will essentially handle traffic from all high-volume data centers (host sites such as the Center for Academic Computing). The outer ring is a Proteon ProNET-10, which is a 10 megabit/second star-shaped fiber-optic token ring network. The outer ring will handle traffic between the lower-volume sites (user PC labs, departmental LANs etc.). See Figure 1 on the following page for a layout diagram. The ProNET-80 and ProNET-10 are high-performance modular network systems which provide, in one architecture, interfaces to a wide variety of local network types and a range of interfaces to available host types. 2.4 Protocol Transmission Control Protocol / Internet Protocol (TCP/IP) is the leading open, non-vendor-specific telecommunications software protocol family available today. It is widely used in government, scientific and educational applications. It is comprised of individual protocols for file transfer, remote login, mail transfer and network maintenance. It is the standard for the Defense Department, the National Science Foundation supercomputer projects and many universities. TCP/IP permits communication not only within but also between networks. Gateways allow interoperable connections between IEEE 802.5 (IBM Token Ring), ProNET-10, ProNET-80 and IEEE 802.3 (Ethernet) type local networks. Products are available for support of TCP/IP on a variety of hosts and a growing number of LANs. TCP/IP is implemented as the campus backbone network protocol directly in the backbone gateway/router nodes. It is implemented in the campus host sites accessible through the backbone either as host-resident software or as code in front-end processors. In the local networks attached to the backbone, it exists in the form of special software customized to the local network or of firmware present on a given network board. 3. CONNECTION 3.1 Physical Connection Use of the high-speed network is dependent on physical backbone connections. These connections consist of physical optical fiber links, fiber-driver units and router-gateway nodes. Arrangements about physical link cost and installation scheduling must be made in advance with OTC. LANs must be of a type supported by the router-gateway interfaces. 3.1.1 Fiber Links University Park's conduit system radiates from a central point, the Telecommunications Building, in star-like fashion to virtually all major buildings on campus. Fiber optic cable has been pulled through the conduit to many of these locations. It is necessary to have a fiber link before a backbone network connection is possible. OTC has formal policies on inter-building and intra- building cabling which specify design, installation and maintenance considerations as well as identify how the costs are apportioned for any given connection. OTC policy statements on these subjects can be found in Appendix 4 of this document. In order to generate a connection as described, the potential user must contact OTC directly for information on availability and installation scheduling. 3.1.2 Gateway Access Points User access to the high-speed data backbone is by means of an interface to a network gateway/router. The network gateway is a high performance modular internetwork router which makes possible the integration of a variety of local and wide-area networking facilities. It offers multiple network interface support within a single gateway. The gateway operates under its own packet-switching software which controls the behavior of the different forwarding and routing protocols, provides for fail-safe operation and monitors network performance statistics. 3.1.3 Local Network Attachment Attachment of the local network to the backbone will be made via a cable from the gateway (supplied by OTC) to an interface (supplied by the user) on the local network. The following network types are supported: * IEEE 802.5 (IBM Token Ring) * ProNET-10 * ProNET-80 * IEEE 802.3 (Ethernet) * Appletalk (Ethertalk or Localtalk via a Localtalk to Ethernet gateway) Each of these network types is implemented by installation of the corresponding network interface board in the router/gateway and by attachment of the appropriate cable from the gateway to the user interface. 3.2 Protocol Implementation In order to communicate over the high-speed network once a physical link is operational, it is necessary to implement at least a minimum set of the TCP/IP communications protocols. There are different LAN strategies for networking PCs and/or terminals but all of them must meet this basic requirement. Examples of successful existing PSU LAN implementations illustrate this. 3.2.1 Features of TCP/IP TCP/IP is a family of related protocols. A minimum set of these protocols must be implemented in order to operate at Penn State in the data communication environment. These include: * File Transfer Protocol (FTP) -- allows file transfer between hosts. It supports text and binary file transfer. Files can also be listed, deleted and renamed remotely. All operations take place under password control. * Telnet Remote Login Protocol -- allows users on one machine to login to another, just as if they were using a directly connected terminal. Through one terminal or PC any user can login to any host under password control. This particular protocol is used in two major forms at Penn State: TN3270 (Telnet in 3270 mode) which gives access to IBM hosts (PSUVM etc.) and TNVT (Telnet in VT100 mode) which gives access to most other campus host sites such as ECL. * Trivial File Transfer Protocol (TFTP) -- is a simplified file transfer protocol, often used for network bootstrap. Additional features may be represented in any given implementation of TCP/IP such as: * Domain Nameserver Protocol -- permits central storage of names used to address host systems etc. Names are consistent and meaningful, even across multiple networks. * Internet Control Message Protocol (ICMP) -- provides functions for maintenance and routing control. * Berkeley Extensions -- include a number of additional protocols originally developed as part of Berkeley UNIX TCP/IP. These protocols provide additional flexibility in communicating with hosts running 4.2 BSD UNIX or derivatives. * Programming Interfaces -- are available for most TCP/IP implementations for developing applications programs that communicate over the network. 3.2.2 PCs In order for devices on a local network to communicate with devices on other local networks via the backbone, they must use the TCP/IP protocol suite. For PCs running on a local network, this will mean either running some PC network software package which supports TCP/IP or using a network board which has TCP/IP in onboard firmware. OTC recommends a TCP/IP network package such as PC/TCP from FTP Software in Cambridge, MA. PC/TCP is written with specific device drivers for different network architectures. Some of the network types supported by PC/TCP software are: * Proteon ProNET-10 * IBM Token Ring * 3Com 3C500 Etherlink, 3C503 Etherlink II, 3C505 Etherlink Plus * Ungermann-Bass NIC Ethernet * Generic Ethernet (customer supplies device driver) * Excelan Ethernet * Banyan VINES Ethernet * MICOM-Interlan Ethernet * Western Digital Ethernet There are obviously different strategies for creating local PC networks which may be compatible with the requirements of the backbone network. Examples of a few already operational local network sites are as follows: OTC maintains its own LAN which ties together all the data facilities of its daily operations. This LAN is a ProNET-10 token ring network. It connects together the Telecommunications Building, which is the main OTC center, and Pine Cottage, where the Data Services are located. The network makes available to the staff a main Banyan File Server running Vines software and several different printer units and is attached through a gateway/router to the backbone network. OTC uses PC/TCP from FTP Software as its network package. The Boucke PC Lab and the Findlay PC Lab, which are operated by the Center for Academic Computing, are ProNET-10 token ring networks of IBM PS/2 Model 30s. Each network includes a Banyan File Server running Vines software and different printers and is attached through a gateway/router to the backbone. TCP/IP communications is implemented with PC/TCP from FTP Software. The College of Human Development Computer Lab (Henderson Building) operates a 3Com Etherlink PC network which includes a 3Com file server and printers and is attached to the high-speed network through an Ethernet interface to a backbone gateway/router. TCP/IP communications is once again implemented via PC/TCP from FTP Software. Potential users interested in LAN implementation should also refer to the document "Networking Microcomputers at Penn State", recently published by the Penn State Office of Computer and Information Systems. 3.2.3 Asynchronous Terminals At the present time serial RS-232 is the Penn State standard for switched terminals and the hosts they access. Because asynchronous terminals attached to local networks would also have to speak TCP/IP in order to access campus hosts over the backbone network, the requirement in this case would be to attach them to the local network with a terminal server speaking TCP/IP. At this time, however, OTC does not recommend that terminal servers be used as a general solution for devices that cannot be networked in the style described in Section 3.2.2, because even though there are terminal servers available that speak TCP/IP protocol, there are none that implement TN3270, which means they cannot access all Penn State hosts. OTC is currently monitoring the major terminal server vendors in order to encourage the eventual development of TN3270 protocol. 4. ACCESS PROCEDURES Connection to the high-speed network gives access to computing data resources both on and off the PSU campus. The backbone network is connected by router-gateways to links to the Internet so that familiarity with the Internet addressing and naming schemes allows the user to specify interactive or file transfer connections to TCP/IP host sites either on or off the campus. 4.1 On Campus Correct attachment to the Data Backbone and correct implementation of TCP/IP protocols allow the local user access to a range of available Penn State computing resources. TCP/IP networks and hosts are accessed by reference to an IP address given as a set of 4 decimal numbers such as 128.118.25.2. The structure of the address gives you some information about how to get to the system you wish to access. For example, 128.118 is a network number assigned by a central authority to Penn State University. Penn State uses the next number to represent campus subnetworks so that 128.118.25 is the OTC Ethernet in the Telecommunications Building. The last number is a host number on the local subnetwork. 128.118.25.2 thus is the address of the cisco terminal server on the OTC Ethernet on the Penn State campus network. Since we normally wish to refer to systems by name rather than by IP address, there is also a TCP/IP name structure for system addresses. The name is given as a set of subnames such as PSUVM.PSU.EDU or SHIRE.CS.PSU.EDU. Each of these subnames represents a separate Internet "domain". EDU, for example, represents the domain of educational institutions. PSU.EDU is, once again, a centrally assigned name designating Penn State University (PSU) as an educational entity. CS.PSU.EDU designates the Computer Science Department subnetwork at Penn State and PSUVM and SHIRE represent host names. TCP/IP networks then make use of "nameservers", which are host computers that translate Internet names into correct IP addresses. Most TCP/IP networks also make use of "domain nameservers" which translate names only for a given domain and thus allow the use of shortened Internet names such as PSUVM for PSUVM.PSU.EDU. In order to access Penn State computing facilities a user must then designate the TCP/IP service desired and either the appropriate host facility IP address or its equivalent Internet name. For example: > TN3270 PSUVM.PSU.EDU establishes an interactive session to the Center for Academic Computing IBM facility. This could be condensed to TN3270 PSUVM since the Penn State domain name server would resolve the short name. > FTP PSUVM would allow uploading and downloading of files between the user's PC and the IBM. > TNVT ECL.PSU.EDU establishes an interactive session to the Electrical Engineering Computer Lab VAX cluster. Again, this could be shortened to TNVT ECL. > FTP ECL would establish file transfer connection to the same site. There are many other host facilities available than the examples named here. It is not the purpose of this document to attempt to enumerate all such facilities. Other campus host services which still rely on RS-232 connections through data switches may be accessed from the backbone network by connecting via TNVT to the cisco terminal server which is maintained by OTC to give backbone attachment to the IDX Data Switch. The cisco terminal server is connected to groups of IDX RS-232 ports which are directly attached to predesignated host services. These services can be accessed by TNVT to the following IP addresses: 128.118.25.11 IDXE (IDX switch, 7 bits, even parity) 128.118.25.12 IDXN (IDX switch, 8 bits, no parity) 128.118.25.13 LIAS 128.118.25.14 CC1 128.118.25.15 CC3 128.118.25.16 DEC10 (ECL) 128.118.25.17 PEN 128.118.25.18 LEPS For example, TNVT 128.118.25.13 will connect the user through the IDX data switch to the LIAS library access system. One note on the use of IDXE and IDXN: after receiving the screen message that the specified connection has been made, the user may enter a "?" in order to review a list of host services accessible from these connections. The above-mentioned use of the cisco terminal server IP addresses is also useful in one other Data Backbone networking situation at Penn State. As previously stated, OTC does not presently recommend the use of terminal server devices in order to attach asynchronous terminals to local networks with access to the Data Backbone because of the inability of these devices to implement TN3270 protocol. However, if a terminal server has the ability to Telnet to IP addresses in general, then it is possible to connect to the cisco services specified above. The specifics of making such a connection will be dependent on the particular terminal server implementation of TCP/IP. The College of Human Development Computing Lab currently operates a Bridge Communications CS/100 Communications Server on its local Ethernet which implements these connections correctly. For all users who wish to attach to PSUVM from computers which do not support TN3270 protocol or from terminals running through terminal servers on local networks attached to the data backbone, OTC has now made an additional connection service available. Users may acess this service by using TCP/IP Telnet (TNVT) to the hostname CAC3270.PSU.EDU. This will connect them to a PSUVM front-end processor running Yale Ascii Terminal Communication System (YTERM). The user will be prompted for terminal type and should reply with the appropriate entry (VT100, HDS etc.). Typing a '?' character at the prompt will display a list of accepted terminal types. One other mode of access to the data backbone remains to be noted. If the user currently has an RS-232 connection to the IDX data switch, he or she may request attachment to OTCTN in response to the IDX prompt and be connected to the cisco terminal server directly. From the cisco it is possible to connect to any regular TCP/IP Data Backbone network host facility by specifying either the IP address or the Internet name of the service desired as described above. 4.2 Off Campus Because the high-speed backbone is attached through a an OTC-managed cisco router to off-campus networks, backbone users also have access to off-campus data resources. Penn State is a member of the Consortium for Scientific Computing, which operates the John von Neumann Center (JvNC) at Princeton. A T1 circuit links University Park to JvNC, through which University users can access NSFNET. With the University's growing interest in supercomputing activities it has also become an affiliate member of the Pittsburgh Supercomputing Center, the National Center for Supercomputing Applications at the University of Illinois and the Cornell National Supercomputing Facility, all of which can be accessed through NSFNET. Penn State is also a member of the Pennsylvania Research Economic Partnership network (PREPnet). PREPnet is a newly formed education and research network established in order to link the major Pennsylvania universities and other associated research centers such as the Pittsburgh Supercomputing Center into a network facilitating high-speed and more direct communication paths among the member institutions. In order to access off-campus Internet locations a user must follow the same procedure described for TCP/IP service to on-campus services. For example, interactive connection to the John von Neumann Center (JvNC) at Princeton would be established by TNVT JVNCC.CSC.ORG. File transfer access to the same location would be established through FTP JVNCC.CSC.ORG. 5. SUPPORT 5.1 Ordering a Backbone Connection Potential users may initiate the process of ordering a backbone connection by contacting OTC. The Data Communications Services Office of OTC is located in Pine Cottage on the University Park campus. The telephone number is 863-1354. A data engineer from the office will assist in the preparation of a Telecommunications Service Requisition (TSR), which specifies where the connection is to be made; what budget, fund and project will pay for the connection; who authorizes the connection for the local user etc.; and what the installation and monthly cost figures will be. Presently the charge for a high-speed host gateway is $3,200.00 and the monthly connection fee is $275.00. Other technical and cost considerations for the potential client are: * Inter-building cabling -- The inter-building cabling used to support public networks is considered as infrastructure and is centrally funded. Special-purpose private networks requiring inter-building cabling are cost-recovered. * Intra-building cabling -- Because the intra-building cabling (such as riser cable) required to connect the gateway to the inter-building cable plant is not funded as infrastructure, OTC has absorbed the cost for this portion of the network. * Demarcation point -- The intra-building fiber-optic cabling is terminated by a fiber-optic interface which connects directly to the network gateway/router. The gateway installation includes the network card for the type of local network and an appropriate cable to connect to the client LAN interface device. This is the demarcation point for OTC's network management considerations between the backbone and the user's local network. OTC policy statements on these considerations are included as Appendix 4 of this document. 5.2 Network User Responsibilities and New User Requirements Each site must designate a person or persons to act as administrator and manager of the local area network that is or will be connected to the Data Backbone. The administrator/manager will be responsible for: * Acting as LAN liaison. The most common times this will be required are: - for the installation of the data backbone equipment and initial connection of the LAN to the data backbone. - at times, maintenance of the data backbone equipment may require the assistance of the LAN administrator. - to serve as the contact person for dissemination of pertinent information. The administrator will inform the LAN users of any changes or events that may effect usage. OTC will keep the administrator informed of any changes. Information will flow in both directions. Mike Contino, 5-0859, mac at psuvm.psu.edu, is the contact person in OTC. Electronic mail is the accepted means of communicating network information. The administrator is almost required to have an account on a network-connected machine and check the mail daily. - in the event of network trouble the administrator should be in contact with OTC in order to coordinate efforts to identify and correct the trouble condition. * Manage the LAN subnet number and assign host IP addresses. * When required, prepare, maintain and make public a table of host names on the LAN. This is only required for those sites with hosts that provide a service. The Engineering Computer Laboratory (ECL) is an example of a site for which this is required. It is required since ECL provides service across the network. A site of only personal computers would usually not be required to make a host table since personal computers typically only provide service to the person at the keyboard and not to network users. When required, a host name table should be maintained on a networked host, either one under local control or one that the local organization can access. If a host table is to be implemented, host and file names should be communicated to OTC. For a layout of host table format, see Appendix 1 of this document. 5.3 Reporting Data Network Trouble To report backbone network trouble, such as degradation or interruption of service, the user should call the OTC Trouble Line: 863-1254. APPENDIX 1 HOST TABLE FORMAT The purpose of the host table is to provide the official Penn State name servers with the name/address combination of publicly available hosts. The entry for a given host is comprised of a few records. The records describe the name and address of the host, the cpu type and operating system type, the servers the host provides, mail delivery point and any number of records for aliases or nicknames. Notational Conventions: anything in < > is the name of an argument. In an actual entry a specific name or value replaces the < >. For every host in the host table, the entry would look like the following, which we will refer to as our example: IN A
IN HINFO IN WKS
[] IN MX and any number of optional alias records of the form: IN CNAME The 'A' record data is the Internet address in dotted decimal form. A sample might look like: OTC1 IN A 128.118.25.25 The 'HINFO' record gives information about a particular host. The data is two strings separated by whitespace. The first string is a hardware description and the second is software. The hardware is usually a manufacturer name followed by a dash and model designation. The software string is usually the name of the operating system. Official HINFO types can be found in the Assigned Numbers RFC, the latest of which is RFC-1010. Following the example given above: IN HINFO SUN UNIX The WKS record is used to list well-known services or host providers. The WKS record lists what services are available at certain addresses using certain protocols. The common protocols are TCP or UDP. Official protocol names can be found in the Assigned Numbers RFC. Following the example started above: IN WKS 128.118.25.25 TCP TELNET FTP SMTP MX records specify where mail for a domain name should be delivered. Currently every host receives its own mail. Following the above example: IN MX 5 OTC1 The CNAME record is used for nicknames. Nicknames are optional and there may be more than one. For example: OTCNM IN CNAME OTC1 Bringing together all the example records from above, a complete entry for a single host is a host table would look like: OTC1 IN A 128.118.25.25 IN HINFO SUN UNIX IN WKS 128.118.25.25 TCP TELNET FTP SMTP IN MX 5 OTC1.PSU.EDU OTCNM IN CNAME OTC1 The host table would have one entry for each host. The host table is used to provide information to the Penn State name servers. Reference information for name servers and the format of data records is in: RFC 1010 Assigned Numbers RFC 1032 Domain Administrator's Guide RFC 1033 Domain Administrator's Operator's Guide RFC 1034 Domain Names - Concepts and Facilities RFC 1035 Domain Names - Implementation and Specifications All these documents are publicly available via anonymous FTP from the Internet Network Information Center (NIC). File name format for the RFCs is: rfc:rfcxxxx.txt and they can be retrieved using the following approximate dialogue: >ftp sri-nic.arpa --- --- Name: anonymous --- Password: username --- ftp> get rfc:rfcxxxx.txt --- --- --- ftp> quit APPENDIX 2 PSU NETWORKING GLOSSARY Backdoor connections Method used by owners of private networks to provide access to other organizations. This method is usually due to the use of proprietary protocols on the private net. Traffic from users on the private net that is destined for other organizations first goes through some intermediate system before reaching the public network. Data Backbone OTC-owned and operated network used to interconnect local networks owned and operated by other organizations. The network operates at the OSI Network Layer. Today it is based on the TCP/IP standard routing protocol, IP. Private network A network owned and operated by an organization for its own purposes. Does not use the Backbone for inter- building connections. Generally, does not provide paths to other organizations except through "backdoor" methods. Public network A network owned and operated by OTC that provides access to all Penn State public services. Public services Computing services that are available to all Penn State personnel. Some examples are the Center for Academic Computing (CAC), LIAS and PENpages. Regional hub Project to provide the state network with more bandwidth, reduce circuit cost, better network management and displace data switch ports. State-wide network OTC-owned and operated network used to interconnect locations outside of the State College area with University Park and each other. Can be used as part of the public network or used as an intercampus carrier facility for private networks. Switched network Network made of asynchronous devices attached to data switches. The switches are "loosely" connected by an OTC data switch. Generally, an "originate" device directly attached to the OTC data switch is considered part of the public network while those attached to other switches are not. Tail circuit Term that refers to the section of a link that runs from the user's device to the location where the major carrier facility terminates. Most commonly a dedicated copper circuit with line drivers that link a device to the state-wide networking components at a campus. APPENDIX 3 GENERAL NETWORKING GLOSSARY Copyright (c) 1986, 1987, 1988 by FTP Software, Inc. All rights reserved. The following informal definitions may help the inexperienced reader to understand networking and our package better. Application Layer An ISO layer that performs services for the user, for example, file transfer. ARP The standard Address Resolution Protocol (for Ethernet only) specified in RFC 826 which resolves Internet addresses into Ethernet hardware addresses. ARPA The Advanced Research Project Agency, an agency of the U.S. Defense Department. Also referred to as DARPA, the Defense Advanced Research Project Agency. ARPANET A network of computers, located principally at U.S. universities (but including connections at other types of sites both in the U.S. and abroad), in use since the 1960s and sponsored by ARPA. Bandwidth The data rate of a device or link. Bridge A device used to expand a local area network (especially an Ethernet) by forwarding packets between data link layers. Usually, a bridge's behavior will initially be identical to a repeater, but as it learns about the network, it will decide whether or not to forward a packet based on what it has learned. Also known as a smart repeater, a smart level 1 forwarder, or a LAN bridge. Broadband A transmission system in which signals are applied to the transmission medium after being translated in frequency. This allows the system to handle many simultaneous signals or be distributed alongside existing signals (for instance, over a CATV system). Broadcast A message system in which all messages are heard by all hosts. Bus Network A network topology in which messages are broadcast to all hosts over a full duplex medium. Ethernet is a bus network. CCITT The Comite Consultatif International de Telegraphie et Telephonie. An international standards organization similar to IEEE, which sets international communications usage standards. Client A computer which is configured to request services from a network. Client can also refer to the code at one end of the network connection. This end of the connection is also referred to as the user end. Collision Detection The process of detecting that simultaneous interfering transmissions have taken place. Typically, each transmitting host that detects the collision will wait for some period of time and try again. CSMA/CD Carrier Sense Multiple Access with Collision Detection. A method of transmission across multiple hosts on a bus using collision detection to determine if more than one signal is present on the bus at any given time. CSMA denotes that you listen to the transmission medium before sending any information. Ethernet is a CSMA/CD network. DARPA See ARPA. Datagram A method of transmitting messages in which sections of the message are allowed to be transmitted through the transmission system in scattered order, and the correct order is reestablished by the receiving host. Datagram usually implies unreliable delivery. Data Link Layer An ISO layer which provides host-to-host delivery across a LAN. This layer performs the assembly and transmission of messages, including error and flow control procedures. DDN The Defense Data Network. A network which provides long-haul and area data communications and interconnectivity for DoD systems, and supports the DoD suite of protocols (especially TCP and IP). All equipment attached to the DDN by military subscribers must incorporate, or be compatible with, the DoD internet and transport protocols. DoD The U.S. Department of Defense. It has sponsored a number of projects which have considerably advanced communications. Domain A naming category in the domain naming system. For instance, a host named silk.ftp.com has two levels of domain name. It has a hostname of silk, and is part of the ftp domain within the com domain. Domain Naming System A hierarchial system of host naming. It allows for grouping hosts into categories. For instance, in the ARPANET naming scheme, hosts with extensions of .COM are commercial hosts, and names with extensions of .EDU are educational hosts. Driver A software module that controls an input/output port or external device. PC/TCP uses a device driver to control the network interfaces. Ethernet The IEEE 802 standard for link and physical layer specifications. The 802 family of Ethernet standards includes four combinations of network topologies and access methods, but 802.3 is the topology generally referred to as Ethernet. _____________________________________ Standard Description _____________________________________ 802.3 A bus using Carrier Sense Multiple Access with Collision Detection (CSMA/CD) 802.4 A bus using token passing (token bus) 802.5 A ring using token passing (token ring) 802.6 A metropolitan area network Note that there are subtle differences between 802.3 and the DEC/Intel/Xerox Ethernet specification (also known as Bluebook or Version 1). These differences will become more important as more implementations which conform to the 802.3 specification appear. Finger A standard protocol specified in RFC 742 which lists who is currently logged in on another host. Flow Control Hardware or software mechanisms employed in data communications to turn off transmission when the receiving host is unable to store the data it is receiving. FTP The standard File Transfer Protocol specified in RFC 959. It provides full authentication of the user by requiring login on the remote host. It allows the user to transfer text and binary files to and from the PC, list directories on the foreign host, delete and rename files on the foreign host, and perform wildcard transfers between hosts. Gateway A computer and its associated software that permit two networks using different protocols to communicate with each other. A gateway translates all protocol levels from physical layer up through applications layer, and can thus be used to interconnect networks that differ in every detail. In the TCP/IP community, this is sometimes referred to as a translating gateway. Some people also refer to a router as a gateway. Glass Terminal A keyboard and screen that conveys data generated by the user directly to a computer or network without buffering or otherwise acting upon the data, and also returns data from the computer to the user unchanged. This terminal type does not provide for cursor addressing or escape sequences. Host A computer which acts as a client and/or server. Host name Resolution A mechanism which provides static and dynamic mechanisms for resolving host names into numeric addresses. The Internet Name Server Protocol accesses an Internet Name Server which provides dynamic name-to-number translation (this process is specified in IEN 116). The Domain Name Protocol accesses a Domain Name Server which provides dynamic name- to-number translation (this process is specified in RFCs 882 and 883). A static local host table can also be accessed for name-to-number translations. Host table An ASCII text file where each line is an entry consisting of one numeric address and one or more names associated with that address. Host tables are used to resolve host names into numeric addresses. IBM Token Ring An IEEE 802.5 ring network which uses token passing, sold by IBM. ICMP The standard Internet Control Message Protocol specified in RFC 792 which provides a number of diagnostic functions and can send error packets to hosts. ICMP uses the basic support of IP and is an integral part of IP. IEEE The Institute of Electrical and Electronic Engineers. This body acts as an information exchange, publishing, and standards-making body responsible for many standards used in local area networks, notably the 802 series. IEN An Internet Experimental Note. It is a standards document similar to an RFC available from the Network Information Center (NIC). These documents contain suggestions and proposals for INTERNET implementations or specifications. Internet Layer An ISO network layer protocol that provides host-to-host delivery across an internetwork. It only provides for transmitting blocks of data (datagrams) between hosts identified by fixed length addresses. There are no mechanisms found in host-to-host protocols to augment end-to-end data reliability, flow control, sequencing, or other services. Internet Protocol (IP) is one of the protocols used in this layer. Internetwork Networks linked together to form a larger network; a network of networks (as opposed to a network of hosts). INTERNET The network running Internet Protocol across the United States and Canada. It includes the ARPANET and consists of over 30,000 hosts. IP The standard Internet Protocol specified in RFC 791 which provides for transmitting blocks of data (datagrams) between hosts identified by fixed length addresses. There are no mechanisms found in host-to-host protocols to augment end-to-end data reliability, flow control, sequencing, or other services. ISO Layer One of the seven layers of the International Standards Organization (ISO) Open Systems Interconnection (OSI) reference model. A table of the ISO layers with examples of implementations for the IBM PC follows. _____________________________________ ISO Layer Implemented As _____________________________________ Application Network applications, such as PC/TCP programs Presentation DOS operating system and programs Session NETBIOS Transport TCP, UDP, etc. Network IP, ICMP, etc. Data Link Ethernet, ProNET, Starlan, X.25, etc. Physical Tranceiver, cable, interface board, etc. The network, transport, session, and presentation layers each provide a set of functions. These functions may be delegated to different layers by different protocol families and vendors. Local Area Network (LAN) A high-bandwidth network spanning a small area (e.g., buildings) over a (usually) inexpensive medium owned by the users. Link Layer Access Method The algorithm that determines when any given network interface is allowed to transmit. It is also known as the access method. CSMA/CD is the access method for the Ethernet. Mail Server A host and its associated software that offer electronic mail reception and (optionally) forwarding service. Users may send messages to, and receive messages from, any other user in the system. Medium An electronic pathway, or mechanism for passing information from one point to another. Metropolitan Area Network (MAN) A communications network that spans a larger area (for example, a township) which provides high-bandwidth communication over moderately inexpensive media, and is provided to all users for a subscription fee. MILNET A network which provides long-haul and area data communications and interconnectivity for DoD systems, and supports the DoD suite of protocols (especially TCP and IP). It is a subset of the INTERNET. NETBIOS NETwork Basic Input/Output System. It provides a Session Layer interface between network applications running on a PC and the underlying protocol software of the Transport and Network Layers. It supports Name Resolution, Datagram and Session services. Network Address A unique number associated with a host which identifies it to other hosts during network transactions. NIC Network Information Center. This is a clearinghouse for network information. They provide network numbers for the Internet, and they collect and disseminate information on protocol standards, such as IP and UDP. Network Layer An ISO layer that receives data from the data link layer and passes the data (after this layer has performed it's services) up to the transport layer. Nicname A protocol specified in RFC 812 which requests information about a specific user or host name from the ARPANET Network Information Center (or NIC) name resolution service. Physical Layer An ISO layer that involves the electrical process of getting the data from one point to another. Ping A program which is useful for testing and debugging networks. It sends an ICMP Echo packet to the specified host, and waits for a response. It reports success or failure and statistics about its operation. Port As used here, a transport-user identified. For example, on UNIX hosts running SMTP, port 25 is reserved for SMTP interactions with other hosts. Presentation Layer An ISO layer that receives data from the session layer and passes the data (after this layer has performed it's services) up to the application layer. ProNET-4 A 802.5 token ring network sold by Proteon Inc. ProNET-10 A proprietary star-shaped token ring network sold by Proteon Inc. Protocol A set of rules for communications. Protocol Layers A way of thinking about communication protocols that models them as a hierarchial family in which each protocol obtains services from protocols beneath it and performs a service for protocols above it. The division of the hierarchy are referred to as layers or levels. Repeater A device used at the physical layer that amplifies or otherwise conditions signals received from one piece of a transmission medium and passes them on to another, similar piece of transmission medium without reading or altering the addresses or the data content. Retransmission A method of error control in which hosts receiving messages acknowledge the receipt of correct messages and either do not acknowledge, or acknowledge in the negative, for the receipt of incorrect messages. The lack of acknowledgement, or receipt of negative acknowledgement, is an indication to the sending host that is should transmit the failed message again. RFC A Request For Comment document from the Network Information Center (NIC). These documents contain suggestions and proposals for network implementations or specifications. Ring Network A uni-directional network topology in which each host passes all messages on, after copying any messages addressed to it. The IEEE 802.5 network and the ProNET-10 are both ring networks. Router A device that receives physical layer signalling from a network, performs data link layer and network layer protocol processing upon those signals, and then sends them via appropriate data link layer and physical layer protocols onto another network. The information for the layers above the data link layer remain unchanged. The primary function of a router is to determine how to forward a packet toward its destination, based on a system within the router that indicates the cost, congestion status, and other factors associated with possible routes. Also called a level 2 packet forwarder. In the TCP/IP community, this is sometimes referred to as a gateway. Routing The decision making process which determines what path a packet takes as it transverses a network. Routing may be based on tables which indicate the cost, congestion factors, and other factors associated with possible routes, or it may be based on a static routing table. Server A computer which is configured to provide services to a network. Server can also refer to the code at one end of the network connection. Session Layer An ISO layer that receives data from the transport layer and passes the data (after this layer has performed its services) up to the presentation layer. SMTP The Simple Mail Transfer Protocol specified in RFC 821 which is used to send and receive electronic mail. Socket As used here, a synonym for port. Spooler A program which queues input for later output. For example, a print spooler can accept files at a high transfer rate, then send them to a printer at whatever rate that printer can handle. Starlan An implementation of the 802.3 Ethernet network in which hosts are connected via telephone-style wiring, which is sold by AT&T and other vendors. Star Network A network topology in which the central node knows the path to all of the other nodes. Using this topology, it is very easy to control access to the network and perform network maintenance and testing. Subnet A single LAN within an internetwork or a collection of LANS. TCP The standard Transmission Control Protocol specified in RFC 793. TCP is a transport layer, connection-oriented, end-to-end protocol providing reliable, sequenced, and non-duplicated delivery of bytes to a remote or local user. TCP provides reliable byte stream communication between pairs of processes in hosts attached to interconnected networks. Telnet The standard TCP/IP remote login protocol specified in RFC 764. Using Telnet, you can work from you PC as if it were a terminal attached to the other machine via a hardwired line. TFTP The Trivial File Transfer Protocol specified in RFC 783. It provides a subset of the functionality of FTP with no user authentication. Token A unique combination of bits, which, when received, indicates permission to transmit. Token Bus The use of a token to control access to a bus in a token bus network. A host receiving the token is allowed to transmit, then it passes the token on to a specified next address. Token Ring The use of a token to control access to a ring in a token ring network. A host receives all messages currently circulating on the ring, followed by the token. The host may allow everything to pass and append a message to those circulating, and then reinsert the token. Trailer A non-standard way of sending data specified in RFC 893. They are used on some networks by Berkeley 4.2 and 4.3 BSD Unix and some of its derivatives. See the section entitled "Trailers" in the chapter entitled 'Problem Solving'. Transport Layer An ISO layer that provides user-to-user communication. It receives data from the network layer and passes the data (after this layer has performed it's services) up to the session layer. UDP The standard User Datagram Protocol specified in RFC 768. UDP is a transport layer, connectionless mode protocol providing a (potentially unreliable, unsequenced, and/or duplicated) datagram mode of communication for delivery of packets to a remote or local user. It provides a procedure for a process to send messages to other processes with a minimum of protocol mechanism. Whois The Whois protocol is an option of the Finger protocol that requests additional information. It is specified in RFC 812. X25 A standard network protocol for the data link layer that was standardized by CCITT, and adopted for use by the Defense Data Network to use in Interface Message Processor (IMP) communication. APPENDIX 4 RELEVANT OTC POLICY STATEMENTS Office of Telecommunications Policy On Backbone Attachments September 14, 1987 The purpose of this policy is to define the interfaces between the data backbone and a user's network, and to identify the different funding sources involved in supporting the interconnection of the two. The backbone includes all devices and cabling necessary to support data transmission between the interfaces defined below. It includes devices such as gateways and wiring centers, any inter- and intra-building cabling (and the associated conduit space) and software needed to enable communications between the gateways. A local network is one which provides data communications among user locations independently of the existence of the backbone. Local network components include devices such as Ethernet transceivers, DELNI's, Pronet wiring centers, and other equipment used for local area networking, as well as intra- (and, in some cases, inter-) building cabling. Attachment of the backbone to the local network will be made via a cable from the gateway (supplied by the OTC) to an interface (supplied by the using office) on the local network. The local network interface could be a transceiver on an ethernet, a wire center on a proNET, or a DESTA on a "thin-wire" ethernet. The point where the gateway cable meets the interface becomes the demarcation point. Everything from that point outward is the responsibility of the OTC. Everything from that point inward is the responsibility of those operating the local network. Diagram A (see following page) shows the demarcation points in the physical and logical perspectives. Because the intra- building cabling (such as riser cable) required to connect the gateway to the inter-building cable plant is not funded as infrastructure, OTC has previously absorbed the cost for this portion of the network. However, future plans include recovering costs of this cabling by assessing a one-time charge to the user, based on actual time and materials required. The gateway will continue to be cost-recovered by charging a one-time and a monthly charge for its use. The cabling outside the building is subject to the provisions of the policy on inter-building cabling. The physical connection is not all that's needed. In order for devices on the local network to communicate with devices on other local networks over the backbone they must use the TCP/IP protocol suite. Although this protocol is not new, its use at Penn State is. Because of this, OTC personnel may be heavily involved in both acquisition and implementation of hardware and software for local networks requiring attachment to the backbone. The amount of support needed will vary from organization to organization and will generally diminish over time, as the support of the local network is ultimately the responsibility of the organization operating the local net. As a service, OTC provides ongoing advice to organizations requesting it. Two areas that may need continued support are the interaction of different TCP/IP implementations and enhancements to existing packages, particularly PC software. Office of Telecommunications Policy on Routing DECnet Packets on the High-Speed Data Backbone June 20, 1988 The high-speed data backbone is the public data communications network that links together computers owned and operated by various Penn State organizations. As such, it supports the only standard data communications protocol available today that allows this connectivity, TCP/IP. The backbone itself is constructed from TCP/IP routers located at end user sites and fiber optic cables that link the sites together. At the end user site, the TCP/IP router is interfaced to the user's local area network. (See the University policy on backbone attachments for more information on this interface.) The functionality provided by the TCP/IP routers is to move packets to and from end user sites over the fiber-optic network. As a result of the efforts of the manufacture of our TCP/IP router, we now have the capability to route packets of a non- standard protocol over the backbone. That protocol is DECnet, a data communications protocol suite developed by the Digital Equipment Corporation (DEC). A few Penn State organizations would like to have this capability supported for their own private use. There are advantages gained by doing this, but someone will have to bear the costs of adding additional software to the routers. Our long-range plan for the backbone is to begin migrating it to the OSI protocol suite as soon as practical. It is likely that the backbone will support both TCP/IP and OSI during the transition phase. DEC, the computer manufacturer whose Penn State customers are requesting DECnet on the backbone, has repeatedly and publicly made a strong commitment to the OSI protocol suite. In view of DEC's commitment to OSI, OTC will temporarily support the routing of DECnet packets over the backbone with no additional connection charges. Once the backbone is capable of routing OSI packets with at least the same degree of functionality as offered by TCP/IP today, and DECnet Phase V (the version of DECnet that will be OSI compatible) is publicly available, DECnet will continue to be supported for one year. After that time support for DECnet on the backbone will be discontinued. Office of Telecommunications Policy on Intra-building Cabling September 14, 1987 The purpose of this policy is to define responsibilities relating to the specification, installation and maintenance of intra- building communication cabling. It is the responsibility of OTC to establish standards for intra- building cabling. These standards are to be based on proven technologies and serve the majority of user applications. Since the networks using this cabling generally serve more than one user group and more than one floor of a building, OTC is responsible for establishing minimum standards for intra-building wiring. However, it is not the intent of this policy to address casual installations, but rather to establish the scope of authority in cases of new construction and renovations, and establish installation standards for user requested installations. For all installations it is recommended that OTC be consulted before cable is purchased or installed. In those cases where OTC is requested to install cabling, OTC will install the cabling on a time and material basis, unless it is included as part of a plan funded as infrastructure. In those cases where the cabling is not in accordance with the standard, or where the installation is of a piecemeal nature, the installation will be referred to the Office of Physical Plant (OPP). Likewise, where requests are received for the installation of user-supplied cabling, they will be forwarded to OPP, regardless of whether the cabling is in accordance with the standard or not. OTC will provide maintenance for any cabling installed by the Office. Such maintenance will include repair or replacement as deemed necessary by OTC personnel to restore working service. Support will also include testing for verification of acceptable performance with new services. Services that cannot be provided by standard wire/cable and are not public networked services may require the user to install private wire/cable. OTC will not provide maintenance of customer-supplied cabling. Office of Telecommunications Policy on Inter-building Cabling September 14, 1987 The purpose of this policy is to define responsibilities relating to the specification, installation and maintenance of inter- building communication cabling. There are two categories of inter-building cabling--that which is used to support public networks (as exemplified by the data backbone and by the video distribution system), and that which is used to support private networks (as exemplified by particular linkages between mainframes or using unsupported protocols.) For the public network case, it is OTC's responsibility to set standards and request central funding for the installation of the cabling, based upon the current perspective that such cabling should be considered as infrastructure. In these cases, the cable will be installed according to a schedule approved by the Telecommunications Advisory Committee, who will also approve any changes to the schedule or the priorities listed. In this case, OTC is responsible for the design, installation, and maintenance of the cable. In cases where a private network requires inter-building cabling which can be installed in accordance with both the technical specification and the schedule, the user will be billed at a standard rate per foot, based on the amount of fiber used. This rate may include a prorated amount for the use of the conduit space consumed. In the case where the private network requires installation of cabling at a time inconsistent with that of the schedule, or requires a different type of cabling than that supported by OTC or is installed via a different route, the installation shall be at OTC's discretion, with billing based on a time and materials basis for the number of cables requested. The use of approved cable types is strongly encouraged. For installations of non- standard cable types, OTC will coordinate installation by OPP. After installation, OTC is responsible for allocating the use of all telecommunications cable. OTC must authorize the use of conduit space for all communications cable.